I have made a video on svi 2 pro taake a loook at it and all the necessary links are in the description of the video
Description
FAQ
Comments (17)
Can anyone help with svi node, i have a wanimagetovideosvipro node missing and unable to install
I have the same problem. :(
trying to install via comfy manager didn't work for me, so i had to install by cloning https://github.com/kijai/ComfyUI-KJNodes directly into custom_nodes.
@Morgennes in isntalled directly using terminal and it fixed it
Same problem for me, seem the manager do not update two files: nodes.py and initpy, I've downloaded them and updated manually.
@superacf135 howd you update manually?
@zimzaderk_ deleted the folder and used terminal to clone the github repo
Simply update comfyUI to latest version and reinstall the node
what about tv2?
Tested on a 3080 10GB and it runs great. I’ve attached a sample video. 👌
Looking for some help. I've tried this on multiple workflows with SVI. No matter what I do with the pro LORAs, the SVIPro nodes in KJNodes, or edit steps and CFG, I get nothing but noisy videos. Anyone know the cause?
For anyone else dealing with this, it has to do with the KJNodes. I have the node pack isntalled with the new WanImageToVideoSVIPro node installed but comfyui is not seeing or using it. I'm in the ComfyUI Desktop version. It shows installed but it's not functioning. More to follow...
Hi, does anyone know why, when I generate videos at resolutions higher than 480px, only a few seconds of the video are generated and the rest is just noise, or is the entire video noise? I'm using a Civitai model from WAN2.2 gguf, which isn't the base model. Could that be the problem? During generation, I get this error: "Tried to unpin tensor not pinned by ComfyUI." I don't know if that's the issue. I hope someone can help me.
check you arent mixing T2V and I2V models. if you are using an input image you need to use I2V
@Burning_Dark Thanks for your reply, hehe. I only have I2V, I don't use T2V, so I'm sure of that. That's why the noise is so strange to me.
Thanks for the workflow. First one I've tried that is working well for me.
I'm getting an odd result where every video segment seems to start with the start image, so the final output doesn't make much sense. I had turned off sage attention, but I wouldn't expect that to be the cause of the issue here